Character Design
Building a character from the numbers up
A method for starting a role-playing character by listing what the sheet will be asked to do, then paying for it — rather than writing a backstory and hoping the dice cooperate.
Most character creation advice starts with a story. A tabletop session does need a story, but the sheet in front of you is also an interface: a list of numbers that will be read aloud at moments of pressure, and those numbers decide how often your character gets to do the thing you imagined.
This note offers a method that starts from the numbers and works outward to the story. It is our own suggestion, not a rule from any rulebook, and it ends with a sheet that does not argue with itself.
Step one: list the jobs
Before spending a single point, write down the five things you expect this character to attempt regularly. Be specific and be honest about frequency. A list that looks like this is more useful than “good at fighting”:
- Talk a guard into letting the group pass.
- Climb or balance during a chase.
- Notice an ambush before it starts.
- Keep a wounded companion on their feet.
- Hit a moving target with a thrown weapon.
Each job is a skill check waiting to happen. A character who attempts four of these every session and one of them once a campaign should be built for the four. That sounds obvious until you compare the list with the sheet of a player who spent points on the character they wish they were playing.
Step two: rank the jobs by frequency
Put the five in order, most attempted first. This ranking, not taste, decides where points go: a point is worth more on a job you attempt often.
If two jobs happen equally often and one is more dangerous to fail, break the tie in favour of the dangerous one. Failing to climb a wall leaves you on the ground; failing to stop a wound from bleeding is not the same kind of failure.
Step three: know what a point is worth
Here is where arithmetic replaces intuition. In a roll-under system on a twenty-sided die, one point of skill is worth exactly 5% on a plain roll — one face out of twenty. Raising a skill from 8 to 12 adds four faces, and the chance goes from 40% to 60%.
That linear step is worth keeping in mind. In the middle of the range, each point is worth the same as the last. Near the ceiling, points run out of room, because the die cannot show a result above 20. A character who has pushed a skill to 18 has only two faces left to gain, and at 20 any further point is wasted; a point spent there is a point not spent lifting the third-ranked job from 9 to 10.
Step four: budget by tier, not by totals
A budget we suggest, to be adjusted to your own system’s point costs:
- Primary job: one skill above the midpoint, enough that a routine attempt usually succeeds.
- Secondary jobs: two skills around the midpoint, so a trained character is reliable without being cinematic.
- Coverage: a little training in two skills that will come up in emergencies, such as noticing, resisting, or getting out of trouble.
- No-everything: whatever is left stays unspent until the first three sessions, because real play reveals which job you underestimated.
Character creation often happens before anyone knows how often each job will come up in the campaign. Holding a few points back lets the first sessions correct the list.
Step five: the story that fits the numbers
Now write the backstory, and let it explain the distribution you just built. A character with a high primary skill, two mid skills and emergency coverage reads as someone who trained hard at one trade, worked steadily at two more, and survived enough close calls to have learned the value of a quick eye.
The reverse is what usually goes wrong: a backstory claiming mastery of five disciplines, followed by a sheet that supports none of them, and a player who spends the campaign frustrated because the fiction promised what the numbers never delivered. Working numbers first produces a backstory that the sheet is already prepared to act out, which is a quieter pleasure than a dramatic pitch and lasts far longer.
A worked example
Suppose the jobs are: talk to guards, climb, notice ambushes, patch wounds, throw knives, in that order. The table below uses a roll-under d20 for simplicity (the free rules use roll-over against a Difficulty Class, which gives the same chances with the numbers mirrored). The targets and attempt counts are illustrative figures we chose for the example, not data from a real campaign.
| Job | Rank | Roll-under target | Chance on one d20 | Attempts per session (assumed) | Expected successes | Value of one more point |
|---|---|---|---|---|---|---|
| Talk to guards | 1 | 14 | 70% | 6 | 4.2 | 0.30 |
| Climb or balance | 2 | 13 | 65% | 4 | 2.6 | 0.20 |
| Notice an ambush | 3 | 10 | 50% | 3 | 1.5 | 0.15 |
| Patch wounds | 4 | 6 | 30% | 2 | 0.6 | 0.10 |
| Throw a weapon | 5 | 5 | 25% | 1 | 0.25 | 0.05 |
The last column is attempts × 0.05, because each point adds one face out of twenty. Over the 16 assumed attempts the sheet expects 9.15 successes, about 57%. Moving four points from the top job to the fourth job (talk 14 to 10, patch 6 to 10) lowers the expected total to 8.35, because those four points were worth 0.30 each where they started and 0.10 each where they landed. That is the case for ranking before spending.
Two jobs are reliable, one is a coin toss, two are there for emergencies. The third job will sometimes decide an evening, and those moments will be tense because the character is not built for them. A player who knows where the character is weak can plan around it and ask the table for help, which is what a party is for.
What this method is not
It does not tell you which skill list is best, because that depends on the system, and systems differ too much for a single note to cover. It does not promise a powerful character; it promises a coherent one. And it does not replace reading the rulebook, because the point costs and the die type change the arithmetic more than any method can.
What it does is make the first decision honest: what will this character actually be asked to do? In the free rules, uncertain actions are decided by a d20 roll compared with a target number (D&D Beyond — Free Rules: Playing the Game), and the chance of any outcome is a probability over a set of possible results (OpenStax — Mean or Expected Value and Standard Deviation). Spending points without knowing the answers to those two questions is spending them blind.
Method
Chances are face counts divided by 20; expected successes are attempts multiplied by chance. The reallocation figure is the sum of the table’s expected successes with the top job lowered by four faces (6 attempts × 0.2 = 1.2 lost) and the fourth job raised by four faces (2 attempts × 0.2 = 0.4 gained). The numbers were checked by script. The rules description was read on 2026-10-31.
Sources
- D&D Beyond — Free Rules: Playing the Game: the d20 Test, with a roll compared against a target number.
- OpenStax — Mean or Expected Value and Standard Deviation: outcomes and their probabilities, and the expected value as a probability-weighted sum.
Written for: a player building a character for a long campaign who wants the sheet to match the play instead of the fantasy.
Sources
- D&D Beyond — Free Rules: Playing the Game (d20 Tests and ability checks) — accessed 2026-10-31
- OpenStax — Mean or Expected Value and Standard Deviation — accessed 2026-10-31